VMware Explore 2026: Broadcom tackles AI server DRAM crunch with NVMe memory tiering

At VMware Explore 2026, Broadcom put forward a technical answer to one of AI infrastructure's most pressing problems: the AI server DRAM crisis. The solution is NVMe memory tiering.

As Tech Times reports, tight DRAM supply has become a bottleneck for large-scale AI deployments. Broadcom's approach uses NVMe storage as part of a memory tiering scheme, expanding the pool of usable memory beyond DRAM.

Memory tiering itself is not a new idea, but in the AI server context its value is clear: systems can gain near-DRAM capacity at a much lower cost, easing the pressure created by scarce high-end memory modules.

The move fits Broadcom's positioning in the VMware ecosystem. Since acquiring VMware, Broadcom has kept combining virtualization and infrastructure software with hardware solutions, and memory tiering extends that strategy into AI infrastructure.

For data center operators, NVMe memory tiering means more memory headroom for AI training and inference workloads without relying on expensive DRAM expansion.

The broader signal: DRAM shortages are pushing infrastructure vendors to explore new memory architectures, and tiered-memory approaches are moving from niche to mainstream discussion.

What to watch next: concrete performance numbers for the scheme, how it integrates with VMware's platform, and whether major server vendors adopt the approach.